Author: Yi-Chang Sung, Andrea Wu

Supervising Editor: Andrea Wu

As the global technology industry competes to build more powerful AI models and greater computing capacity, SoftBank has turned its attention to the physical infrastructure that supports artificial intelligence. According to Reuters and official information released by SoftBank, the company plans to develop 3.1 GW of AI data center capacity in the Hauts-de-France region of northern France. The first phase is expected to involve approximately EUR 45 billion in investment, while the overall project could eventually expand to 5 GW, with total investment reaching as much as EUR 75 billion. This is not merely a corporate data center expansion project. It reflects a broader shift in which AI data centers are evolving from enterprise IT facilities into a new form of infrastructure that integrates computing power, energy, land, and industrial supply chains.

From the moment a user enters a prompt and receives an immediate response, a complex network of GPUs, AI servers, high-speed connectivity, power systems, and cooling technologies is operating behind the scenes. AI services can no longer be viewed solely as software innovation. As demand for model training and inference continues to grow, computing power is increasingly becoming a strategic resource in the digital economy, while AI data centers are emerging as a central arena in the competition for future technological leadership.

1. How Are AI Data Centers Different from Traditional Data Centers?

Traditional data centers primarily support websites, enterprise systems, cloud storage, e-commerce, video streaming, and general computing workloads. Their main function is to provide a stable environment for data storage and service operations. However, with the rapid development of generative AI, large language models, and enterprise AI applications, data centers are increasingly responsible for high-intensity computing tasks such as model training, inference, data processing, and intelligent service deployment.

Compared with conventional facilities, AI data centers rely much more heavily on the coordinated operation of GPUs, CPUs, AI accelerators, high-bandwidth memory, high-speed networks, storage systems, and software platforms. NVIDIA has described this trend through the concept of the “AI Factory,” referring to data centers not simply as spaces for housing equipment, but as a new type of production base that transforms data into intelligence. As a result, computing power is evolving from an IT resource into a strategic asset for both corporate and national competitiveness.

2. Why Are Countries Competing to Build AI Data Centers?

SoftBank’s investment in France has attracted attention not only because of its scale, but also because it reflects a shift in the site-selection criteria for AI data centers. In the past, data center locations were primarily evaluated based on land costs, network connectivity, and proximity to major markets. Today, AI data centers depend more heavily on stable and scalable power supplies, grid connection capacity, cooling conditions, and government policy support.

Reuters reported that SoftBank founder Masayoshi Son viewed France’s energy production and export capabilities as an important factor supporting AI infrastructure investment. This demonstrates that energy conditions have become a key source of competitiveness in attracting data center projects.

Land and site conditions are also becoming increasingly important. Selecting a location for an AI data center is no longer simply a matter of finding available land. Developers must also assess the availability of stable electricity, expandable grid connections, suitable industrial sites, and favorable local regulatory conditions.

In SoftBank’s French project, Schneider Electric is expected to provide modular solutions, while the state-owned utility EDF will offer former power plant sites for redevelopment. This illustrates how AI data centers have become cross-industry projects involving energy companies, equipment suppliers, local governments, and technology firms.

Thermal management and cooling technologies are another critical factor in determining whether AI data centers can operate efficiently. AI servers equipped with high-density GPUs and accelerators generate significantly more heat than conventional servers. Inadequate cooling can reduce equipment performance while increasing energy consumption and operating costs. This means that competition in AI data centers is no longer limited to semiconductor performance; it also depends on energy availability, site conditions, and engineering integration capabilities.

3. How Are AI Data Centers Becoming Tools for National Investment Promotion?

France is not the only country actively seeking to attract AI data center investment. According to official information from OpenAI, the Stargate initiative led by OpenAI, Oracle, and SoftBank has added five new AI data center sites in the United States. Together with the flagship site in Abilene, Texas, and related CoreWeave projects, the planned capacity is approaching 7 GW, with more than USD 400 billion in investment expected over the next three years.

This demonstrates how the United States is accelerating the development of domestic AI computing infrastructure through a combination of major technology companies, cloud platforms, and large-scale private capital.

The Middle East presents a different model, one that integrates energy resources with national strategy. According to Reuters and OpenAI, Stargate UAE is expected to begin operations in 2026, with an initial deployment of 200 MW and a long-term planned capacity of up to 5 GW.

Located in Abu Dhabi, the project is being developed through cooperation between the UAE-based company G42 and international technology firms including OpenAI, Oracle, NVIDIA, Cisco, and SoftBank. It is regarded as one of the largest AI data center clusters outside the United States.

OpenAI has also described the project as the first international deployment of Stargate. In addition to building a 1 GW AI computing cluster in Abu Dhabi, the arrangement includes UAE investment in Stargate infrastructure in the United States. This indicates that AI data centers are no longer merely commercial investments; they are increasingly becoming part of technology diplomacy, sovereign capital strategies, and cross-border cooperation.

In Europe, Spain has emerged as an important location for Amazon Web Services’ expansion of AI infrastructure. According to Reuters, Amazon announced an additional EUR 18 billion investment in Spanish data centers and AI innovation, bringing its total committed investment in the country to EUR 33.7 billion. The company also positioned Spain as one of its major European AI operating hubs.

This case shows that European countries are linking AI data center investment not only to cloud computing and capacity needs, but also to employment creation, regional development, and technological competitiveness.

Taken together, the cases of the United States, the UAE, Spain, and France demonstrate that AI data centers have evolved from internal corporate IT facilities into strategic tools for national investment promotion, technology diplomacy, and regional industrial competition. Countries are competing not only for AI models and GPUs, but also for the electricity, land, policy frameworks, and infrastructure integration capabilities required to support AI development.

4. What Can Taiwan Learn from This Trend?

As AI data center construction accelerates worldwide, Taiwan’s role is not limited to supplying individual components. It has the potential to expand from hardware manufacturing into broader solution-based cooperation.

Taiwan has long-established strengths in semiconductors, AI servers, motherboards, power supply systems, thermal modules, networking equipment, and system assembly. These capabilities correspond directly to the key requirements of AI data centers. As cloud service providers and governments continue to invest in AI infrastructure, Taiwanese companies are expected to benefit from growing demand for AI servers, liquid cooling systems, power architectures, and high-speed networking equipment.

According to the Market Intelligence & Consulting Institute’s review of the ten major technology trends for 2026, global cloud service providers and AI start-ups are planning hundreds of billions of dollars in capital expenditure and actively purchasing AI servers. MIC also estimates that total server shipments will reach 15 million units in 2026, with AI servers accounting for approximately 30%, or about 4.5 million units.

At the same time, rack-scale AI platforms are expected to drive upgrades in cooling and power supply solutions. Advanced liquid cooling and high-voltage direct current power systems are likely to become essential for the stable operation of high-density AI computing platforms.

However, for Taiwan to capture these opportunities more fully, it cannot focus solely on server shipments and supply chain orders. It must also consider the supporting conditions required for AI infrastructure, including electricity, land, cooling, site availability, and policy coordination.

For Taiwan, the central question is not only how much hardware it can supply, but whether it can build on its existing supply chain strengths to participate more deeply in AI infrastructure integration and international cooperation.

5. Conclusion

From SoftBank’s investment in AI data centers in France to the Stargate projects in the United States and the UAE, and Amazon Web Services’ expanded investment in Spain, it is clear that AI data centers are no longer merely internal corporate facilities. They are becoming a new form of infrastructure jointly developed by major technology companies and national governments.

The competition driven by generative AI is no longer confined to models and applications. It now extends to computing power, energy, data centers, and supply chain integration. For enterprises, access to stable and scalable computing resources will influence whether AI can move from experimentation to practical deployment. For governments, AI data centers are becoming important nodes in the next phase of technological competition, energy planning, and international cooperation.

Disclaimer

This article is based primarily on publicly available information and is intended solely for general informational and reference purposes. It does not constitute investment advice, a basis for business decisions, or professional consultation.

The technologies, data, and cases discussed in this article may vary over time depending on their sources and changes in market conditions. The author does not guarantee the completeness or timeliness of the information. Readers should independently assess the relevance and applicability of the information before using it and should assume responsibility for any associated risks.

Reference 

Reuters reported on SoftBank’s planned investment in AI infrastructure in France, including the development of AI data center capacity in the Hauts-de-France region.

SoftBank’s official press release also outlined its AI data center investment plan in France, emphasizing the strengthening of France’s AI infrastructure and Europe’s technological sovereignty.

NVIDIA explained that the Enterprise AI Factory is a validated design that provides full-stack guidance for building and deploying enterprise AI factories, covering interoperability among enterprise AI software, open models, and infrastructure platforms.

OpenAI announced that OpenAI, Oracle, and SoftBank would add five new Stargate AI data center sites in the United States. The announcement stated that the overall planned capacity was approaching 7 GW, with investment expected to exceed US$400 billion over the next three years.

Reuters reported that Amazon would invest an additional €18 billion in data centers and AI in Spain, bringing its total committed investment in the country to €33.7 billion and positioning Spain as a core hub for its European AI operations.

The International Energy Agency stated that global electricity consumption by data centers is projected to rise to approximately 945 TWh by 2030, nearly double the 2024 level, with data center electricity demand growing faster than that of other sectors.

Microsoft explained that its next-generation data center design incorporates chip-level cooling for AI workloads and reduces cooling water consumption through an architecture that does not rely on evaporative cooling.

MIC’s Ten Key Technology Trends for 2026 estimated that total server shipments would reach 15 million units in 2026, with AI servers accounting for approximately 30%, or around 4.5 million units. The report also noted that rack-scale AI platforms would drive upgrades in cooling and power supply solutions.

The National Development Council’s Five Trusted Industry Sectors initiative identifies semiconductors and artificial intelligence as priority industries. It emphasizes the development of the semiconductor, artificial intelligence, defense, security and surveillance, and next-generation communications sectors to position Taiwan as a trusted partner within the global democratic technology ecosystem.